Executive summary

• Between 2010 and 2025 the UK Home Office removed 28,118 people on 730 charter flights to a small set of destinations.
• Over the disclosed window, more than 53,000 deportation directions were issued and 55% were withdrawn before removal.
• Volumes peaked in 2013 (~2,600), collapsed by 74% in 2019 (from ~1,600 to ~410) in the aftermath of the Windrush scandal, recovered through the early 2020s, reached a new peak in 2023-2024 (2,622 / 2,616), and dropped to 1,389 in 2025. TODO[incomplete-2025]
• Charter removals are a minority channel (10-32% of yearly Home Office-arranged returns across 2010-2025), reserved for specific cohorts and specific destinations.
• Albania and Kosovo became the dominant destination from the mid-2010s onward, accounting for 38% of all charter deportations across the period. Together with Pakistan they cover over 55% of the total. Other major destinations are Afghanistan (stopped after 2015) and EU (started growing from 2020 onward).
• EU withdrawal rates dropped 28 percentage points at the Brexit boundary (48% under Dublin III to 20% under bilateral arrangements). The data cannot apportion this gap between contracted legal recourse, bilateral-framework cooperation, and UK-side operational change.
• The same EU regime change reversed the composition of who is removed: under Dublin III roughly 1% of EU deportees were classified as FNOs, under bilateral arrangements roughly 87% are.
• The UK-Albania readmission agreement of December 2022 coincides with monthly Albania-Kosovo removals more than quadrupling (+353%). The agreement formalised an existing 8-year ramp-up: Albania-Kosovo was already ~60% of charter-flight deportees by 2018.
• Women are a small minority of charter-flight deportees (8% across the period), and the few who are removed are concentrated on Sub-Saharan Africa routes. That region accounts for 14% of all deportations but 42% of all female deportations.
• Four 2024 charter flights to Brazil removed ~720 adults (28% of 2024 adult deportees) and were the only flights in the dataset accompanied by children (~130). While the Home Office classified them as voluntary returns under the Voluntary Returns Service, advocacy groups contested the operation as coercive in practice.
• Total disclosed cost: ~£41 million in real 2025 £, across 2020-01 to 2023-06 (charter-contract fees only). Escorting costs are excluded from FOIA disclosure: every £ figure is a lower bound on true public spend. Mean cost per flight is ~£214,000, mean cost per person removed ~£8,800.
• Of the 321 flights (2020-2025) for which use-of-force counts are disclosed, 140 (43.6%) recorded at least one incident. The escort-to-deportee ratio is ~2.1:1.
• Per-flight passenger counts ranged from 1 to 203, with a mean of 39. On several occasions, a charter flight was used to deport a single person.

Introduction

Charter flights are the Home Office’s most expensive, operationally intensive, and politically visible removal channel. Unlike scheduled-flight removals, charter operations require an aircraft for a single removal route, in-flight private-security escorts, and coordination with the receiving state. They are reserved for specific destinations (where bilateral or, until 2020, EU arrangements make group removals viable) and specific cohorts (where the Home Office judges that scheduled-flight removal would not be suitable).

This report assembles, in one place, what the FOIA record actually shows about that activity: where the flights went, who was on them, what happened to those who were issued removal directions, what the state paid, and how often force was used. Analysing the whole series, rather than each FOIA response on its own, surfaces long-term patterns: the evolution in withdrawal rates, the shifts in the destination mix, and the variations across Home Secretaries and major external events.

Terminology. UK law distinguishes between deportation (under the UK Borders Act 2007, typically for foreign national offenders, carrying a re-entry ban with criminal sanctions), administrative removal (under the Immigration Act 1971 or the Immigration and Asylum Act 1999, carrying a shorter bar), and voluntary returns under the Voluntary Returns Service. Charter flights carry a mix of all three, and the FOIA responses report them under one umbrella heading without distinguishing the legal mechanism. Throughout this report, “deportation” is used as a catch-all for every person removed on a charter flight, regardless of statutory pathway. This is a deliberate choice: the government’s semantic distinction between “deportation” and “removal” obscures what is, in every case, the material reality of the state forcibly transporting people out of the country against their will.

Throughout this report, “bilateral arrangements” describes the post-Brexit operational framework for EU removals (direct UK-to-receiving-state coordination, replacing the multilateral Dublin III transfer mechanism that applied until 2020-12-31). It does not imply a network of formal published bilateral return treaties. Only a handful of return agreements with individual EU member states have been publicly confirmed (France, 2025, for small-boat returns; Ireland, 2020, an informal operational protocol for asylum seekers), and the Home Office has not disclosed formal return treaties with most EU countries to which post-2020 charter flights operate.

The data underlying this analysis was reconstructed from FOIA responses obtained by L. Hendry, M. Atkinson, Y. Pahe, and P. Petit between 2014 and 2025. The resulting dataset is one row per flight, with date, destination, total removed, sex breakdown, FNO count, removal directions issued, removal directions withdrawn, and (for post-2020 flights) escort numbers, use-of-force incident counts, and charter-contract costs. The full dataset, parsing code, cleaning rules, and an audit-trail notebook are open source and reproducible end-to-end. 2024 Q1 and 2025 Q2-Q3 data are pending outstanding FOIA requests: figures for those years are incomplete.

Methodology

The analysis is descriptive: it only reports what happened, not why. Where formal methods are used, they attach uncertainty to descriptive statements (whether a trend is real or noise, whether a shift around a policy date is large relative to background variation, etc.): the data and methodology at our disposal do not allow causal attribution, hence we cannot claim that a given policy caused a given outcome.

The substantive findings in this report rest on three methodological choices:

• Per-region rather than pooled trends for withdrawal rates, female share, and FNO share: per-destination trends often move in opposite directions and pooled analysis would average them away
• Composition versus within-destination effects are separated for the three headline rates (withdrawals, female share, FNO share), to display how much of each era-to-era change is driven by the Home Office’s choices of where to fly versus what happened on flights to each destination
• Trend shifts at known policy dates (end of Dublin III, UK-Albania readmission agreement, Home Secretary terms): estimated with before/after regression, and cross-checked against an unsupervised structural-break scan that identifies shift points without event information

Cost figures are reported in real 2025 £, chained from disclosed nominal amounts via annual ONS CPI inflation rates.

The audit notebook (code/deportation_charter_flights.ipynb) carries the full specification, test statistics, sensitivity analyses, and a consolidated Limitations section. This report cites the substantive insights, and any observation can be reproduced or contested within the notebook.

Findings

Charter flights vs scheduled flights

Charter flights as a share of Home Office-arranged returns from the UK, defined as enforced returns plus assisted voluntary returns (Home Office returns statistics, 2010-2025). 2024 and 2025 charter-share figures are understated because charter-flight data for those years is incomplete.

For the period 2010-2018 charter flights were 10-17% of Home Office-arranged returns from the UK in any given year, with a notable dip in 2019 (~4.5%, the broader collapse in charter activity that year) and a jump in 2020-2021 (~20% rising to ~32%) as scheduled-flight removals were curtailed under Covid border restrictions while charter activity rebounded faster than the rest of the enforcement system. From 2021-2023, charter flights accounted for 26-32% of Home Office-arranged returns, roughly double their earlier minority status. The 2024 and 2025 figures (~17% and ~7%) are downward-biased on both sides: charter-flight data for those years is incomplete, and the denominator grew sharply because the Home Office’s assisted voluntary returns programme expanded from ~1,300 returns in 2021 to ~10,300 in 2025. The denominator here is “enforced returns plus assisted voluntary returns”: the Home Office categories that together capture both forced removals (including HO-facilitated returns from detention) and voluntary returns where the Home Office pays for the flight, which is where charter flights classified as voluntary (such as the Brazil 2024 operations) are recorded.

Thus, changes in charter-flight volume reflect changes in how a specific subset of removals is carried out, not the overall scale of enforcement.

Scale and trajectory

The total number of charter-flight deportations falls into four distinct phases:

1. Peak era (2010-2013): Total deportations peaked at around 2,600 in 2013.
2. Gradual decline (2014-2018): A drop to roughly 1,600 by 2016, then a plateau through 2018, driven primarily by reductions in Pakistan and Afghanistan flights.
3. Collapse (2019): Activity dropped to around 410 deportations in 2019, the largest single-year change in the series.
4. Recovery and reorientation (2020-2024): Activity recovered to 2,622 (2023) and 2,616 (2024), but around a fundamentally different destination mix (Albania-Kosovo and EU). 2025 partial figures stand at 1,389. TODO[incomplete-2025]: missing data for Q2-Q3 pending outstanding FOIAs.

Per-flight passenger counts ranged from 1 to 203 (a 2024 Brazil flight). Several flights in the record carried a single person.

Yearly deportations, 2010-2025, by population (total, males, females, FNOs). Female removals followed the overall trajectory. FNO removals stagnated 2014-2018, dropped in 2019, and have grown in line with the overall total since.

Monthly deportations with a 7-month centred moving average, 2010-2025. The smoothed line makes the 2013 peak, the 2015-2018 plateau, the 2019 collapse, and the post-2020 recovery legible.

The 2019 collapse

The sharpest discontinuity in the data: a 74% drop in charter-flight volume between 2018 and 2019 (from ~1,600 to ~410 deportations), with monthly flights falling from a 2010-2018 average of ~7 per month to just 1-2 per month throughout 2019. The drop coincides with the Windrush scandal aftermath: the scandal broke in April 2018 and triggered the Home Affairs Committee inquiry, the National Audit Office investigation, and the Williams Review. Through late 2018 and into 2019 the “compliant environment” measures were partially paused and re-assessed, enforcement operations underwent internal review, and the Home Office’s returns apparatus faced sustained parliamentary scrutiny. An unsupervised structural-break scan independently nominates early 2019 as a break across multiple regional series at once (all flights, Pakistan, Sub-Saharan Africa). The FOIA data cannot distinguish between a deliberate operational pullback and a collateral slowdown in case processing as staff and leadership attention were absorbed by the scandal response, but the co-occurrence is too close to ignore.

The post-2020 system is structurally different: Albania-Kosovo and EU corridors, rather than the Pakistan / Sub-Saharan Africa / Afghanistan mix of the earlier era.

Where the flights go

Ten destination regions appear in the charter-flight record: Albania and Kosovo, Pakistan, Sub-Saharan Africa, Jamaica, Afghanistan, Sri Lanka, Iraq, Vietnam, the EU, and a residual Other category that is dominated, post-2020, by the four Brazil flights.

Total deportations by region, 2010-2025. Albania and Kosovo account for 38% of the total. With Pakistan added, the two destinations cover over 55% of all charter-flight deportations across the period.

Map of total deportations by country of destination, 2010-2025. The geography of charter-flight removal is narrow: a corridor through the Western Balkans, a small number of long-haul routes to South Asia, Sub-Saharan Africa, and the Caribbean, and a wider EU set from 2020 onward.

The destination mix is not static:

• Pakistan was the largest destination by total deportees through the 2010s, peaking in 2013, then collapsed alongside the broader 2019 drop and has not recovered
• Sub-Saharan Africa, one of the main destination throughout the 2010s, disappeared from the dataset after the 2019 collapse, similar to Pakistan
• Albania and Kosovo’s share rose steadily from 2014 onward, accounting for roughly 60% of charter-flight deportees by 2018 (48% of flights, since Albania-Kosovo flights typically carry above-average passenger loads), and consolidating as the dominant corridor from 2020
• Flights to Afghanistan, the main destination in the early 2010s, were stopped after 2015
• EU flights only appear in volume from 2020 onward, first under the residual Dublin III framework and from 2021 under post-Brexit bilateral readmission arrangements

Deportations per destination over time, with centred moving average, 2010-2025. The transition to an Albania-dominated mix is the dominant compositional feature of the post-2020 period.

One of the main insights that emerges from the analysis is that the destination composition shift is itself the policy lever. The Home Office decides which destinations its charter flights serve: it should be viewed as an intentional and influential choice, rather than as a confound to be averaged out when analysing the trends. The destination-mix decomposition section below quantifies how much of the era-to-era change in headline rates is driven by this choice.

Who is on the flights

Demographics

The two panels below summarise who is on the flights to each destination: absolute counts of men, women, and FNOs (left), and the female and FNO share of deportees at each destination (right). The female share is small everywhere (8% across the dataset) but unevenly spread across routes, while the FNO share ranges from near-zero to around 80% depending on destination. The two subsections that follow take each in turn.

Demographic composition of deportees by destination. Left: absolute counts of males, females, and FNOs. Right: the female share and FNO share of all deportees at each destination (the male share is omitted, being the complement of the female share). Destinations are sorted by total deportees.

Women

Women represent roughly 8% of charter-flight deportees across the period, but the cohort of female deportees is distributed very unevenly across destinations.

Sub-Saharan Africa accounts for 14% of total deportations but 42% of all female deportations. Eighteen percent of the 3,855 people deported by charter flight to Sub-Saharan Africa were female, a share that has held steady since 2010-2019, and the highest of any region apart from the 2024 Brazil flights. Jamaica and Sri Lanka follow at 12-13%. The 2024 Other row reads as 34% female, almost entirely driven by the four Brazil flights, which carried families.

Proportion of females among the deported, by destination, 2010-2025.

Every Sub-Saharan destination with more than a handful of removals records a female share above the overall mean.

Distribution of deported females among destinations. Sub-Saharan Africa dominates absolute counts as well as per-flight share.

Female deportations more than halved between 2013 and 2016, driven primarily by reductions in Pakistan and Sub-Saharan Africa flights.

Foreign national offenders

The FNO label is a Home Office category recorded as-is in the FOIA responses. We do not have offence type or severity. Per-destination FNO shares vary enormously: roughly 80% of Jamaican nationals deported by charter were classified as FNOs (the highest of any region), against around 13% of removals to Pakistan and just 6% to Afghanistan.

Albania and Kosovo show a rising FNO share over the post-2020 period (above 50% across the period as a whole).

The EU mix reversed at the Brexit boundary: under Dublin III roughly 1% of EU deportees were FNOs, under post-2021 bilateral arrangements roughly 87% are.

Proportion of FNOs among the deported, by destination, 2010-2025.

Distribution of deported FNOs among destinations. Albania-Kosovo dominates absolute counts as well as per-flight share.

What happens to those issued removal directions

The Home Office issues a removal direction to each person it intends to remove on a given flight. Some proportion of those directions are withdrawn before the flight departs, typically because of legal challenge, last-minute representation, receiving-country refusal, or administrative cancellation. The withdrawal rate is therefore a joint function of UK-side operational practice, the receiving country’s cooperation, the legal grounds available to the deportee, and the capacity of legal and advocacy support. The FOIA records the rate but cannot disaggregate the mechanisms.

Over the disclosed window, more than 53,000 deportation directions were issued and roughly 55% were withdrawn. In the 2010-2019 subset alone the withdrawal rate was 59%.

Per-flight withdrawal rate over time. Each point is one flight, the line is the yearly median. The widening spread of per-flight rates after 2019 is partly mechanical: post-2020 flights are smaller, and a flight with only a few directions can land anywhere between 0% and 100%.

The yearly median per-flight withdrawal rate climbs to a peak around 2019 and then drops sharply, as charter operations shift toward the lower-withdrawal Albania-Kosovo and EU corridors.

Monthly withdrawal-rate volatility rose sharply through the late 2010s, as destination trajectories diverged and per-flight sizes shrank. From 2022 onward it partially compressed again as charter operations consolidated around the lower-variance Albania-Kosovo and EU corridors.

Monthly proportion of withdrawals of deportation directions, with centred moving average, 2010-2025. The aggregate shift downward from 2020 onward reflects both the destination-mix shift toward Albania/Kosovo and the EU regime change.

Divergence by destination

Withdrawal-rate trajectories diverge sharply by destination.

Per-destination median per-flight withdrawal rates (2010-2025), with the number of flights each rate is based on

Destinations with less than 10 flights, as well as the residual Other bucket, are not included, as their medians are too noisy or heterogeneous.

Within destinations, withdrawal rates have been rising for Sub-Saharan Africa and Jamaica (roughly +3 percentage points per year), more weakly for Pakistan, and show no clear trend for Albania-Kosovo or Afghanistan.

For the directions-to-removal ratio we observe Pakistan and Sub-Saharan Africa typically requiring 2-5 directions per actual removal, while Albania-Kosovo and EU flights operating at near 1:1.

Per-flight ratio of directions issued to people actually removed, by destination, 2010-2023. Lower values mean issuance converts more directly to removal. Pakistan and Sub-Saharan Africa flights typically issued 2-5 directions per person actually removed, while Albania-Kosovo and EU sit near 1:1.

The EU-Dublin bilateral regime change

The EU provides the best evidence of the effect of policy on withdrawal rates. Splitting the series at the Dublin/bilateral boundary evidences the shift:

EU withdrawal-rate decomposition at the Dublin/bilateral boundary

The level shift at the boundary is −28 percentage points. And on either side of it, the within-era trends point in opposite directions: under Dublin, withdrawal rates were rising year on year, under bilateral, they are falling.

The potential causal explanations of the shift are multiple.

The most concrete confound is the contraction of legal recourse. Under Dublin III, deportees could challenge removal on Regulation-specific grounds (responsibility determination of the receiving Member State, Charter-of-Fundamental-Rights arguments under EU law, the Regulation’s procedural safeguards). Post-Brexit bilateral readmissions do not sit under EU law, so those specific routes have disappeared. Fewer available legal grounds would, on their own, mechanically lower the withdrawal rate without any change in UK-side practice or receiving-country cooperation.

The targeted population also reversed at the same date: under Dublin III roughly 1% of EU deportees were classified as FNOs, under bilateral arrangements roughly 87% are. Dublin transfers concern asylum responsibility (a largely non-criminal cohort), whereas bilateral removals target foreign national offenders. A different population may carry a different propensity for successful challenge independent of the legal regime, so the composition shift is itself a candidate explanation for the rate drop, and coincides, and might be caused by, the regime change.

The volume data does rule out one narrow reading, more selective issuance of directions only for cases expected to succeed: that would show fewer directions, but issuance rose ~10% (1,845 to 2,052) and completions rose ~70% (955 to 1,636). It does not rule out the deportee demographic mix nor any other operational change.

What the FOIA cannot apportion is how much of the gap is the legal-recourse contraction, how much is the population shift, how much is greater receiving-country cooperation under the bilateral framework, and how much is other UK-side operational change.

Destination-mix decomposition

The post-2020 charter-flight pattern shifted dramatically: Albania and Kosovo became the dominant destination, Pakistan and Afghanistan flights largely stopped, EU flights emerged. Many of the headline trends in earlier sections (lower withdrawal rates, lower female share, higher FNO share, smaller per-flight sizes) are partly attributable to where the Home Office flew to, not to changes in the flights (logistics, legal challenges, etc.).

In plain terms, the question is whether the headline averages moved because the Home Office changed which destinations it flies to, or because something changed on the flights to a given place. The decomposition answers it by splitting each era-to-era change into two parts: a composition part (what the change would have been if every destination kept its own 2010-2019 rate and only the mix of destinations shifted to the 2020-2025 pattern) and a within-destination part (everything left over, i.e. genuine change on the flights to a given place). The two add up to the observed change exactly. The era boundary is 2020-01-01, and moving it to 2019-07-01 or 2020-07-01 leaves the split stable.

Destination-mix decomposition for withdrawal rate, female share, and FNO share. Composition (era-1 base) is the change from re-weighting toward the 2020-2025 destination mix, within-destination (era-2 base) is the residual.

Composition is the policy lever: the Home Office decides where to fly. The composition shifts in this dataset coincide with documented policy and external events: Afghanistan stopped after 2015 (an operational decision), Pakistan collapsed in 2019 (alongside Windrush), EU emerged under Dublin and continued under bilateral arrangements, and Albania/Kosovo dominance from 2020 onward coincided with the December 2022 UK-Albania readmission agreement. The composition therefore identifies the specific mechanism through which Home Office choices produced the observed change in headline rates.

Cost

Charter-flight cost figures are disclosed for 2020-2023 Q2 only, and coverage within the disclosed window is itself patchy.

A critical caveat applies to every cost figure in this report: the disclosed amounts cover charter contracts only and exclude in-flight escorting costs. Published reporting on individual operations suggests that escort fees can rival or exceed the charter contract cost on long-haul routes. Every £ figure here is therefore a lower bound on true public spend, not a full cost.

All figures below are in real 2025 £, chained from nominal disclosed amounts using annual ONS CPI inflation rates.

The highest per-person figure in the disclosed window (January 2020, ~£68k per person) covers two flights (Nigeria/Ghana with 7 people and France with 2 people), with the near-empty France leg dominating the per-person average. The lowest (May 2020, ~£2,500 per person) was a pair of Albania flights carrying 98 people between them. Cost-per-person reflects destination mix: higher Albania share is associated with lower cost per person. Two non-exclusive mechanisms are consistent with this: Albania flights are short-haul and quick to turn around (lower per-flight cost), and Albania flights typically carry more people per flight (higher per-person denominator).

Charter-flight cost per person and per flight by reporting period, 2020 to mid-2023 (real 2025 £).

Use of force on board

The FOIA record contains per-flight use-of-force (UoF) incident counts and escort numbers from January 2020 onward. Pre-2020 has no UoF data. TODO[incomplete-2025]: 2024 Q1 and 2025 Q2-Q3 remain pending.

Across the per-flight window (322 flights, 10,510 people removed):

• 227 total UoF incidents.
• 43.6% of UoF-disclosed flights (140 of 321) recorded at least one UoF incident.
• Escort-to-deportee ratio: 2.1 escorts per person removed (escort-disclosed rows only).

Use-of-force incidents per charter flight, 2020-2025 (monthly average overlaid in red). Most flights record zero UoF incidents. When force is reported, it is typically 1-2 incidents per flight, with occasional larger counts on flights to non-Albania destinations.

The share of flights carrying at least one UoF incident varies considerably across destinations, even before adjusting for flight size:

Share of flights with ≥1 use-of-force incident, by region, 2020-2025, with 95% confidence intervals. Bar opacity reflects how many flights the rate is based on: faint bars (e.g. Pakistan, 3 flights) carry much less weight than solid bars (e.g. Albania/Kosovo, ~170 flights).

These numbers should be read with two caveats. First, the FOIA does not define what constitutes a “use of force incident”. The counts are not person-level and not severity-typed: a single recorded incident may refer to a single person, a single act of force, or a single event involving multiple people and multiple acts of force. Second, the post-2020 destination mix shifted significantly within the UoF window: 2020-2022 covered a wide spread of EU countries, Vietnam, Pakistan and Sub-Saharan Africa, while from late 2022 onward the mix is Albania-dominated. Cross-period comparisons of the monthly UoF rate inherit this compositional change.

Escort-per-person and UoF counts are jointly observed at the per-flight level: the scatter below is the closest the FOIA permits us to get to the in-flight enforcement intensity per individual.

Escorts per person removed versus use-of-force incidents per flight, 2020-2025.

Several 2024 Q1-Q3 rows report escorts = 0, which likely reflects non-disclosure rather than absence of escorts. Per-person ratios use only disclosed rows.

Trajectory shifts at policy events

Two policy events in the window provide reasonably clean before/after contrasts. The same shift is visible whether one looks at the descriptive monthly averages, a controlled before/after regression, or an unsupervised algorithm asked to find structural breaks in the series.

UK-Albania readmission agreement, 13 December 2022

Across the three methods:

• Descriptive means: monthly Albania/Kosovo removals more than quadrupled, from 38.2/month before the agreement to 173/month after (a +353% shift, counting only months with FOIA coverage).
• Before/after regression, controlling for seasonality: an immediate level shift of +124/month.
• Unsupervised structural-break detection: an algorithm asked to scan the Albania/Kosovo series for shifts independently nominates December 2022 as a change point.

However, the agreement formalised an existing trend, it did not initiate one. Albania/Kosovo was already roughly 60% of charter-flight deportations by 2018, and the 8-year ramp-up is visible in the pre-2022 series. The agreement created the bilateral framework under which a long-standing operational priority could be pursued at greater scale. The agreement was the instrument that operationalised and streamlined an already-established priority.

Monthly total deportations to Albania and Kosovo, with macro-level events and policy regimes overlaid. The pre-2022 ramp-up is visible, and the December-2022 level shift is the substantive finding.

End of Dublin III / Brexit transition, 31 December 2020

For EU flights, monthly removals jumped from around 9/month under Dublin (mostly sporadic transfers) to around 44/month under bilateral arrangements (covered months only). The before/after regression puts the level shift at roughly +24/month, though for the EU series this estimate is imprecise: the Dublin-era series is sparse and the firmer evidence for the regime change lies in the collapse in withdrawal rates and the near-total reversal of the FNO mix documented earlier, not in the volume regression. The unsupervised scan independently nominates April 2020 as the structural break in the EU series, broadly consistent with the timing.

For both events, the finding is the one-time level shift: a step up to a higher operating level. What the trend does after the step (whether it keeps rising, holds flat, or drifts down) cannot be read off the data, because the uncovered 2024 Q1 and 2025 Q2-Q3 months sit at the right-hand edge of each series and distort any post-period trend estimate. We therefore make no claim about the post-shift trajectory in either direction. TODO[incomplete-2025]

Monthly total deportations to EU destinations, with the Brexit transition end and Dublin III ending overlaid.

Covid border restrictions, March 2020

Covid is an event we have to describe rather than measure with a before/after comparison, because its effect was temporary rather than a lasting change in level: charter operations stopped almost entirely from April through September 2020 as borders closed, then resumed at low volume from October 2020. That is a roughly six-month gap in the record followed by a return to activity. The volume lost during the shutdown is visible in the yearly totals, but a before/after contrast around the March 2020 lockdown would mistake a temporary suppression for a permanent shift.

Unsupervised breakpoint detection

The figure below overlays the policy-regime annotations on the full all-flights series. The same unsupervised scan run earlier on the Albania/Kosovo and EU series, when run across the nine regional series (all flights, Albania-Kosovo, Pakistan, Sub-Saharan Africa, EU, Afghanistan, Jamaica, females, FNOs), additionally nominates a late-2018 / early-2019 break for all flights, Pakistan, and Sub-Saharan Africa simultaneously (the 2019 collapse already discussed above), and a mid-2015 break for Afghanistan (cessation of flights).

All charter deportations over time with policy regime annotations.

Patterns by Home Secretary

The 2010-2025 window covers nine Home Secretaries, from May (the longest-serving by a wide margin, two-thirds of the window) to Mahmood today. Per-Home-Secretary metrics are observational: each term overlaps with macro events that the Home Secretary did not cause (Covid, the Brexit transition, the Windrush scandal) and with events where the Home Secretary has real influence (the UK-Albania readmission agreement, the Rwanda partnership). The two clearest discontinuities in the post-2020 series sit inside individual terms (the Dublin/bilateral switch in Patel’s, the Albania readmission in Braverman’s), but both are visible at the same dates regardless of who held the office. The data cannot cleanly isolate the Home Secretary’s contribution from the broader environment: what follows therefore reports co-occurrences and the context that overlaps each term.

Yearly averages of each Home Secretary’s term, by metric. Aggregate counts during the term divided by term length in years. May’s totals are largest because she served longest, so per-year averages allow comparison across terms of different lengths.

Overall ratios for each Home Secretary: withdrawn-over-issued, female share, and FNO share. These are aggregate ratios across each term, not per-flight figures.

Reading the series across terms

Deportation directions issued over time, by region, faceted by Home Secretary’s term. Macro-level events that overlap each term are annotated.

Total deportations over time, by region, faceted by Home Secretary’s term. The policy-regime context for the Brexit transition, Dublin III ending, and UK-Albania readmission is overlaid.

A handful of term-level observations are genuinely visible in the data:

• The withdrawal rate steps down along regime lines, not personnel lines. The per-flight median sits in the mid-50s under May (54%), Rudd (55%) and Javid (57%), then falls to the low 40s under Patel (46% median, 37% aggregate) and Braverman (41%, 37%). The drop coincides with the EU regime change and the shift to an Albania-dominated mix, both of which began under Patel, rather than with any change of Home Secretary.
• Rudd issued at close to May’s pace in a fraction of the time. On a per-year basis she issued roughly 90% as many directions as May but completed only ~80% as many removals, posting the highest aggregate withdrawal rate of the period (62%) and its highest month-to-month volatility. Her term also records the highest yearly average of FNO removals.
• Javid’s term shows the sharpest demographic shift. The FNO share among those deported jumps to ~36%, well above the low-twenties of May and Rudd. The “automatic deportation” framework gives the Home Secretary substantial discretion over which foreign national offenders are pursued, and civil-society groups have argued its application is disproportionate. His term also spans the 2019 collapse, which the data cannot apportion between him and Rudd’s last months.
• Two headline policy instruments leave no operational footprint. The Rwanda partnership (announced under Patel in April 2022, ruled unlawful in November 2023, re-legislated in 2024) produced four voluntary departures and zero charter flights. It shapes the legal and political context of the post-2022 period without ever appearing in the flight record.

Two specific operations are worth naming for context. The December 2022 UK-Albania readmission agreement, the largest operational shift in the window, fell weeks into Braverman’s tenure and is treated in full in the trajectory-shifts section above. The four 2024 Brazil flights fell under Cooper, the first Labour Home Secretary in the window: roughly 720 adults, each offered up to £3,000, classified by the Home Office (and Brazil’s Ministry of Foreign Affairs) as voluntary returns under the Voluntary Returns Service and contested as coercive in practice by advocacy groups, who cited the absence of Portuguese-language legal advice, post-Brexit immigration-status confusion, and the removal of children with long UK ties. They are the only flights in the dataset that carried children (~130, counted separately from the adult totals). Mahmood’s term, from September 2025, overlaps only the partial-coverage 2025 data, and substantive observations are deferred until the outstanding FOIAs land. TODO[incomplete-2025]

What the data does not show

The findings above are bounded by what the FOIA responses disclose and by what the per-flight unit of observation can support. The most important limitations:

• No legal-outcome data. The FOIA records withdrawal counts but not why directions were withdrawn (successful judicial review, last-minute representation, receiving-country refusal, administrative decision). The mechanism behind the withdrawal rate is not decomposable from the FOIA alone.
• Use-of-force definition. The FOIA does not define “use of force incident”. Counts are not person-level and not severity-typed and should not be compared to person-level force statistics.
• No cost or legal-aid spend after 2023 Q2. The cost analysis is bounded by the disclosed window and excludes escorting fees throughout. CPI is a generic deflator and may understate aviation/services inflation in 2022-2023.
• Regional grouping. Regions are analytic, not geopolitical. Sub-Saharan Africa flattens 11 countries.
• Cancelled flights. 16 cancelled flights are disclosed in 2020-2025 (~3%). 2010-2019 likely has a disclosure gap on cancelled flights.
• Children are reported only in the Brazil flights. Pre-2020 FOIAs explicitly state no children were carried, 2020-2023 FOIAs are silent and we assume zero. The Brazil flights are the only rows where children are enumerated.
• The “<5” redaction. FOIA responses redact values of 1-4 in small-cell counts. We impute “2” (the midpoint). Sensitivity analysis shows insignificant effects across headline claims.
• FNO classification is broad and opaque. “Foreign national offender” is a conviction-based label covering any non-British citizen convicted of a criminal offence, from minor offences to the most serious. The statutory automatic-deportation duty applies only above a 12-month custodial sentence (UK Borders Act 2007), and below that threshold deportation is discretionary on “conducive to the public good” grounds. The boundary is further blurred by the criminalisation of migration itself, so that offences such as illegal arrival or possession of false documents can now produce the conviction that brings a person within the FNO category. Conviction status is itself downstream of policing and charging decisions: who is classified as an FNO is partly determined by who is charged, and with what. The FOIA do not report offence type or severity, so we cannot see where in that range a given cohort sits.
• Home Secretary attribution is observational only: external events are confounders.
• TODO[incomplete-2025] 2024 Q1 and 2025 Q2-Q3 are pending. Sites in this report that quote 2025 yearly totals or describe a 2024 → 2025 drop will be updated when those FOIAs land.

The notebook code/deportation_charter_flights.ipynb carries the consolidated Limitations section in full, with sensitivity analyses for each major claim.

Conclusion

Across fifteen years the charter-flight programme collapsed and was rebuilt around a different shape. The long-haul, high-withdrawal system of the early 2010s, which had to withdraw half of the deportation directions it issued, fell away in 2019. The programme that recovered after 2020 flies shorter routes (Albania and Kosovo, the EU), and withdraws far fewer directions. The return to 2013-era volumes by 2023-2024 happened around a new geography and a new legal architecture, mainly the UK-Albania readmission agreement and post-Brexit bilateral arrangements with EU countries that anchor the post-2020 corridors.

That reorientation is the key to reading the headline numbers. The largest single driver of almost every aggregate shift in the data is a change in where the Home Office chooses to send people: lower withdrawal rates, a higher FNO share, and smaller flights all follow in large part from the destination mix. Where the within-destination picture does move sharply, at the end of Dublin III, the most concrete candidate is a contraction of the legal grounds on which a removal can be challenged, alongside a near-total change in who is removed (a roughly 1% FNO share under Dublin against roughly 87% under bilateral arrangements).

None of this is cheap, and little of it is gentle. For the window the Home Office disclosed, charter contracts alone cost at least £41 million in real terms, a figure that excludes the in-flight escorting that is often the largest cost on a charter operation. Force was recorded on more than two in five of the flights for which counts exist, escorts outnumbered the people they removed by roughly two to one, and on several occasions an entire aircraft was chartered to remove a single person. Nor are the people on board a cross-section: women, a small minority throughout, are concentrated on the Sub-Saharan African routes, and the only children in the entire record are those flown to Brazil in 2024 on an operation the Home Office classed as voluntary and advocacy groups described as coercive.

Almost everything in this report exists only because individuals filed freedom-of-information requests over more than a decade, and the gaps in what came back are part of the finding. The Home Office does not disclose why a removal direction is withdrawn, what a recorded use of force actually involved, what escorting costs, or what becomes of anyone once the plane has landed. For the questions the record can answer (who was sought, sent where, at what cost, under what force, and how often the state succeeded) the answers are now public, and reproducible in full in the accompanying notebook. Our future disclosure and analysis efforts will attempt to close these gaps.

What should change

Almost every gap in this report is one the Home Office could close. The data assembled here exists only because individuals spent a decade filing freedom-of-information requests and piecing the responses together: none of it should have required that. We ask the Home Office to:

• Publish charter-flight data as routine official statistics. Per-flight records (date, destination, numbers removed, directions issued and withdrawn, demographic breakdown) should be released on a fixed quarterly schedule, not painstakingly extracted one FOIA at a time. The coverage holes that force parts of this report to carry caveats exist only because disclosure is reactive.
• Report the full cost of each operation. Disclosed figures cover charter contracts only and omit in-flight escorting, which can rival or exceed the contract on long-haul routes. Every cost in the public record is therefore a lower bound. Total operational cost, escorts included, should be published.
• Publish why removal directions are withdrawn. More than half of all directions in this dataset were withdrawn before departure, but the record gives only the count, not the reason (legal challenge, last-minute representation, receiving-country refusal, administrative cancellation). Without it the headline withdrawal rate cannot be interpreted.
• Define and report use of force consistently. The FOIA records a bare count with no published definition, no severity, and no person-level detail. A standard definition and severity-typed, person-level reporting would make the figures meaningful and comparable across flights and over time.

These transparency asks would allow the public to know about deportations carried out in its name.

References

Primary data: FOIA responses

The Home Office FOIA responses below were obtained by independent requesters and are bundled with the project repository in data/foia/. Each response is the source of one or more flights in the parsed dataset.

• FOIA 53475 (Hendry). Charter-flight deportations, 2010-2013 (per-flight data with directions, withdrawals, sex, FNO).
• FOIA 57818 (Pahe). Charter-flight deportations, follow-up window.
• FOIA 58660 (Atkinson). Charter-flight deportations, complementary window.
• Petit FOIA series (P. Petit, 2020-2025): quarterly per-flight data for each year/quarter from 2020 Q1 through 2025 Q4 (modulo the outstanding 2024 Q1 and 2025 Q2-Q3 responses).

A full list of every bundled FOIA PDF (filenames, the date ranges they cover, and the cleaning rules applied to each) is available in the docstring of code/parse_foia.py.

Reference data

• UK Home Office, “Immigration System Statistics, returns summary”, December 2025 release. File: data/home_office/returns-summary-dec-2025-tables.ods, Ret_01 sheet. Used for the charter-share-of-all-enforced-returns anchor, covering 2010-2025.
• UK Office for National Statistics, annual CPI inflation rates, 2020-2025. Used to chain disclosed nominal cost figures to a real 2025 £ base. Source rates: 2020 0.9%, 2021 2.6%, 2022 9.1%, 2023 7.3%, 2024 2.5%, 2025 3.4%.
• Natural Earth, 1:50m countries dataset. Bundled as data/countries-50m.json. Used for the map of deportations by country of destination.

Legal and policy instruments referenced

• UK Borders Act 2007, sections 32-33 (automatic deportation of foreign criminals).
• Immigration Act 1971 and Immigration and Asylum Act 1999 (administrative removal powers).
• Regulation (EU) No 604/2013 (“Dublin III”). Ceased to apply to the UK on 31 December 2020 (end of the Brexit transition period).
• Agreement between the Government of the United Kingdom and the Government of the Republic of Albania on the readmission of persons residing without authorisation, signed 13 December 2022, London. The bilateral framework under which post-2022 Albania charter flights operate.
• Migration and Economic Development Partnership between the UK and Rwanda, announced April 2022, and ruled unlawful by the UK Supreme Court in R (AAA and others) v Secretary of State for the Home Department [2023] UKSC 42. Subsequently legislated for in the Safety of Rwanda (Asylum and Immigration) Act 2024. Four voluntary departures took place. No charter flight operated.
• Williams Review, Windrush Lessons Learned Review, HC 93, March 2020.

Appendix: source data and reproducibility

The dataset, parsing code, and notebook are open-source and reproduce end-to-end from the source FOIA PDFs:

uv sync --extra analysis

uv run python code/parse_foia.py  # regenerate dataset

uv run jupyter notebook code/deportation_charter_flights.ipynb

• Source FOIA PDFs: data/foia/ (53475, 57818, 58660 for 2010-2019, Petit_*.pdf for 2020-2025, FOI2023/06120 for the authoritative 2020-2023 use-of-force and escort window).
• Parsed dataset: data/deportation_charter_flights.csv (generated through code/parse_foia.py, except for 2010-2019 data).
• Reference data: data/home_office/returns-summary-dec-2025-tables.ods (Ret_01 sheet), for determining proportion of charter flights as a share of enforced returns, 2010-2025.
• Country geometries for the map: data/countries-50m.json.

Contact: contact@dataactivists.org